MiniMax-M3
Official · BF16 · safetensors
Runtime certified: vLLM SGLang Transformers KTransformers
| Precision / quantization | Format | Bits | Publisher trust | Deployment state |
|---|---|---|---|---|
| BF16 | safetensors | 16 | Official | Runtime certified |
MiniMax · General · MiniMax M3
MiniMax M3: Specs, GPU Requirements and License
MiniMax M3 is a current General release in the MiniMax family, certified from its immutable official source revision.
Total parameters428.0B
Active parameters23.0B
Context window1,048,576 tokens
Release statusCurrent · Current
ArchitectureMiniMaxM3SparseForConditionalGeneration
ModalitiesMultimodal
Licenseother
Deployment stateDeployable with certified configuration
Commercial stateUnknown / Not Recommendable
Release date2026-06-02
Architecture, strengths and use cases
MiniMax M3 is an official self-hostable MiniMax release in the MiniMax General track. It uses MiniMaxM3SparseForConditionalGeneration, carries 428B / 23B MoE, supports a verified context window of 1,048,576 tokens, and exposes only source-certified checkpoints and runtime combinations. Its release, license and deployment facts are tied to the official repository revision rather than inferred from the family name.
MiniMax M3 uses MiniMaxM3SparseForConditionalGeneration with 428B / 23B MoE. The stored architecture profile records 60 layers, hidden width 6144, 64 attention heads and a 1,048,576 token context ceiling where the official configuration publishes those fields.
Verified MiniMaxM3SparseForConditionalGeneration architecture
1,048,576 token certified context
Official BF16, MXFP8, Unknown checkpoint coverage
Private LLM inference
Agentic workflows
Long-context document processing
Certified checkpoints and quantizations
Only exact, source-certified variants attached to this release are shown.
MiniMax-M3-MXFP8
Official · MXFP8 · safetensors
Runtime certified: Transformers vLLM SGLang KTransformers
| Precision / quantization | Format | Bits | Publisher trust | Deployment state |
|---|---|---|---|---|
| MXFP8 | safetensors | 8 | Official | Runtime certified |
Recommended hardware
Use only the listed vllm, sglang, transformers, ktransformers runtime configuration with the exact checkpoint precision. The recommendation engine checks weight memory, KV cache, runtime overhead, safety headroom, GPU architecture and topology before ranking rentable and reference hardware.
Unknown / Not Recommendable
Commercial rights are not verified, so this release is not commercially recommendable.
License and self-hosting considerations
The full 428B resident weight set must fit across the selected GPUs. Context and concurrency add KV-cache memory; MoE active parameters do not replace resident-weight memory.
MiniMax M3 is published under other. The release-scoped official license link and verification date are retained with the catalogue record.
Unknown / Not Recommendable. Commercial rights are not verified, so this release is not commercially recommendable.
Alternatives and sibling variants
MiniMax M2.7, MiniMax M2.5, MiniMax M2.1
Release timeline
minimax/minimax/minimax-m3
Source details
Source revision: 379adf789f3179a77837f6818a8b981c05678577dd940948d3bfc980bc622f4d
Frequently asked questions
How much VRAM does MiniMax M3 require?
The exact estimate depends on checkpoint precision, context and concurrency; the page recommendation uses the certified memory engine and compatible runtime configurations.
Which runtimes support MiniMax M3?
Certified upstream runtime evidence is available for vllm, sglang, transformers, ktransformers.
Can MiniMax M3 be used commercially?
The stored commercial classification is Unknown / Not Recommendable. Consult the official other terms for the final obligations.
Official references
Last verified: 2026-09-03